In order to better understand the impact of environmental policy measures, previous measures can be examined as part of causal analyses. For example, the effects of agri-environmental measures on the environment or farms have been analysed in the past (e.g. review by Uthes and Matzdorf 2012; Mennig and Sauer 2020; Stetter, Mennig, and Sauer 2022; Uehleke, Leonhardt, and Hüttel 2024). In contrast, knowledge on the economics of agriculture in Natura 2000 sites is limited (exceptions are Koemle, Lakner, and Yu 2019; Ahlvik and van Kooten 2024; Thiermann and Bittmann 2023; Grupp et al. 2023; Lakner, Zinngrebe, and Koemle 2020). Natura 2000 forms a network of protected areas in the European Union based on the Flora-Fauna-Habitat (FFH) Directive (92/43/EEC) and the Birds Directive (79/409/EEC), with which the EU aims to improve the protection of valuable species and habitats. In addition to its pure function as a protected area, the designation of Natura 2000 sites also includes sustainable human use (Tsiafouli et al. 2013); across the EU, around 40% of Natura 2000 sites are located on agricultural land, most of which are subject to production restrictions and are used extensively. These extensively used areas often have a high diversity of rare plants and animals, but are often at risk of intensification or abandonment (European Commission 2018). With the adoption of the Biodiversity Strategy 2020, the European Commission is aiming to expand the current protected areas to at least 30% of the land area and 30% of the marine area by 2030. Currently, around 18% of the EU land area and 8% of the water area are protected by Natura 2000 designation, although there is great heterogeneity between the EU member states (8% in Denmark, 38% in Slovenia). Austria, with 15% of its land area (12,895 km2), spread over 352 Natura 2000 sites, is in the middle of the field (Figure 1).
Figure 1: Natura 2000 sites in Austria (as of 2024) by site type (SPA = Special Protected Area; FFH = Flora-Fauna-Habitat, SCI = Special Conservation Interest; data source: EEA)
Previous studies show different economic effects of the designation of Natura 2000 areas. Koemle et al. (2019) show with an analysis of regionally aggregated data with Continuous Treatment Propensity Score Matching (Imbens and Hirano 2004) that the designation of Natura 2000 areas on agricultural land leads to a reduction in rental prices, which indicates a reduced profitability of these areas. In addition, results of an analysis of vegetation indices and night lights at EU level found no significant effects of protected areas on the development and vegetation conditions of European regions (Grupp et al. 2023). In contrast, heterogeneous effects on the value of land were found in a difference-in-differences study in Finland (Ahlvik and van Kooten 2024). The measured effects of Natura 2000 on economic processes are therefore complex, heterogeneous and dependent on the respective context and method of analysis. In order to better understand the economic effects of Natura 2000 on farms, an empirical analysis is carried out in this project. The causal effect of Natura 2000 on efficiency is estimated by combining classical efficiency analysis with causal analysis methods (Figure 2).
Figure 2: Flowchart of the analyses of the proposed project.
Efficiency of farms
In economic theory, the assumption of optimising behaviour (e.g. maximising profit, revenue or the quantity produced) forms the basis of farm management. For example, companies try to maximise profit (in the long term) with given resources. In the productivity and efficiency literature, a distinction is made between the maximum possible (‘efficient’) profit/revenue and the respective amount realised by the company. The difference between the two is referred to as (in)efficiency (Kumbhakar and Lovell 2003). A distinction is made between technical efficiency and allocative efficiency. Technical efficiency describes the ratio between the realised (point XR in Figure 3) and the maximum quantity that can be produced, given the inputs (point XTE in Figure 3). For example, company X is more efficient than company Y if company X manages to produce more output than company Y with the same resources and under the same production conditions. Allocative efficiency in output describes the optimality of the output mix as a function of prices (given complete technical efficiency). In an output-orientated view, both values are between 0 and 1, where 1 means that the company is completely efficient.
Figure 3: Production possibilities and efficiency of a farm with two outputs
Measuring efficiency
Empirically, the efficiency of farms can be measured using stochastic (econometric - Stochastic Frontier Analysis, SFA) and deterministic methods (Data Envelopment Analysis, DEA) (Coelli et al. 2005), both input- and output-orientated. SFA requires the estimation of the respective efficient frontier by defining a functional form of the production, profit or cost function. If several outputs are produced, it is also possible to estimate the dis-tance function, with which the distance from the frontier is modelled directly. The deterministic method of DEA is particularly suitable for splitting the efficiency of a company into technical, allocative and scale efficiency (Pascoe et al. 2022). In addition to the classic DEA, there are now numerous variants with which stochastic elements (e.g. via bootstrapping) can also be incorporated into the deterministic efficiency analysis (Bogetoft and Otto 2010). Although there is already an extensive literature on technical efficiency in agriculture (e.g. Brümmer and Loy 2000; Djokoto 2015; Latruffe et al. 2017; Nowak, Kijek, and Domańska 2015; Thiam, Bravo-Ureta, and Rivas 2001), there is a need for research on the impact of Natura 2000 on the technical and allocative efficiency of farms whose land is located in these areas. The project will analyse the literature to determine the best method (DEA or SFA) for estimating the efficiency of farms located in these areas.
Measuring causal effects
The majority of Natura 2000 sites used for agriculture in the EU are located in marginal locations and include mountain pastures, steppes, heathland and wet meadows (European Commission 2018). The selection of affected farms is therefore not randomly distributed, but particularly affects farms that are at a disadvantage compared to farms in more productive locations. The effects of the designation of protected areas are therefore not directly measurable (e.g. by comparing the average profits between Natura 2000 farms and non-Natura 2000 farms), but require consideration of the respective framework conditions of the farms. The measurement of causal effects has a long tradition in policy evaluation and includes a number of counterfactual methods (Cunningham 2021). The Difference-in-Differences (DiD) method can be used to measure causal effects, for example, without the need to experimentally allocate the ‘treatment’, i.e. the policy measure, from the outset. The literature now includes numerous extensions of the classic DiD method that can deal with panel data structure or interventions introduced with a time lag, as in the case of Natura 2000 areas (Figure 4) (Cunningham 2021; Goodman-Bacon 2021; Callaway and Sant'Anna 2021), or use the advantages of machine learning (handling of high-dimensional data sets, modelling of non-linearity, good adaptation of the models to the data) (Bach et al. 2024; Chernozhukov et al. 2018). In order to reduce the complexity of the control variables (e.g. observations from remote sensing data that describe the heterogeneity of farms and Natura 2000 areas), machine learning is used to summarise the most important components in the preliminary field, which are then included in the estimation of the policy effects.
Figure 4: Number of newly designated protected areas in Austria per year. Cannot be totalled across area types due to double counting (SAC = Special Area of Conservation, SCI = Special Conservation Interest; data source: EEA)
Objective
The main objective of the project is to visualise the production of farms and their technical efficiency in Natura 2000 areas, as well as possible causal determinants of technical (in)efficiency. With the help of the study, the influence of the nature conservation-orientated design of protected areas can be recorded and the need for support in Natura 2000 areas can be better understood.
Detailed objectives
- Expansion and deepening of methodological expertise in efficiency analyses and unsupervised machine learning
- Detailed description of the characteristics of farms that manage land in Natura 2000 areas using FADN and INVEKOS data as well as remote sensing data.
- Description of FFH areas using satellite-based data (e.g. degree of development, vegetation).
- Estimation of efficiency in output and profit, estimation of the heterogeneity of the factors influencing these.
- Formulation of possible policy recommendations for the conservation of agriculture in Natura 2000 areas
- Publication in a scientific journal
Work packages
The project consists of five consecutive work packages and a control package (WP 0).
Work package 0: Project management
Dieter Kömle
The project (project duration: June 2025 to May 2028) includes annual project meetings of the entire team in person or via Zoom, at which the status of the project is discussed, possible decisions are made and interim results are exchanged. Communication also takes place on a bilateral level. At the beginning of the third quarter of 2025, a kick-off meeting will be organised via Zoom for all project partners. The aim is to have a general exchange on the state of knowledge, concretise the research questions and make decisions on the efficient exchange of knowledge, data and literature. Possible hurdles will also be discussed and solutions proposed. Furthermore, dissemination channels outside the scientific literature will be defined. A second project meeting will be organised in mid-2026 to discuss the current status of the project. Here, necessary decisions will be made and the progress of the project in its various areas of work will be presented. At a third project meeting in mid-2027, initial drafts for publications will be presented and final methodological and content-related decisions will be made.
Work package 1: Literature analysis Natura 2000 EU
Dieter Kömle, Yvonne Stickler, Sebastian Lakner
Work package 1 analyses the scientific and political framework of Natura 2000 in the EU. To this end, the scientific literature on the interface between environment/biodiversity conservation and agriculture is systematically collected and analysed; this includes both theoretical papers and empirical studies. In addition, documents from the administration (guidelines, evaluations, etc.) are collected and summarised in order to understand how the implementation of Natura 2000 sites in Austria is situated in an international context.
Work package 2: Natura 2000 and agriculture in Austria
Dieter Kömle
Work package 2 analyses the specific situation in Austria. In addition to a detailed analysis of the various Natura 2000 site types in Austria based on publicly available data (e.g. provided by the European Environment Agency EEA, Federal Environment Agency), this includes initial descriptive analyses of the landscape and agricultural use types of FFH sites. For this purpose, the geometries of the agricultural fields from the INVEKOS database are intersected with the geometries of the Natura 2000 sites. The database on National Designated Areas of the European Environment Agency provides further protected areas designated at national level; these are also included in the analysis where possible. The conservation status of habitats and species is shown in the tables made available by the Austrian Federal Environment Agency. This data is supplemented with information on participation in agri-environmental programmes. Additional data sources are openly available databases such as weather and climate data from Geosphere Austria (e.g. Spartacus), soil data such as the agricultural soil map (Ebod2), vegetation indicators from Sentinel-2 or MODIS satellite data, as well as farm accounting data (FADN). On the one hand, these data are used as control variables in the further analyses; on the other hand, the output of this work package enables a comparison between Natura 2000 areas of the record-keeping farms and the Natura 2000 areas of the remaining INVEKOS farms. With these findings, individual farm results on heterogeneous policy effects, such as the impact of the designation of different types of Natura 2000 areas on the efficiency and productivity of farms, can be extrapolated to the population (Austrian IACS farms).
Table 1: Possible data sources for the present project
Work package 3: Method review and development
Dieter Kömle, Christoph Tribl, (Sebastian Lakner)
In work package 3, the different empirical methods are analysed and compared. Three strands of literature are of importance: Firstly, unsupervised machine learning (clustering) for the classification of Natura 2000 sites based on existing, spatially high-resolution data (e.g. satellite data of vegetation indices, development indicators, description of existing species and conservation statuses, etc.). Secondly, the most important methods for causal inference, and thirdly, the methods for analysing efficiency. In efficiency analysis, there are both stochastic methods (Stochastic Frontier Analysis - SFA) and deterministic methods (Data Envelopment Analysis). The literature analysis will determine which method is most suitable for the proposed project.
Work package 4: Data preparation and analyses
Dieter Kömle, (Yvonne Stickler)
Work package 4 comprises the processing and cleaning of the data collected in work packages 1 and 2 and their analysis using the selected methods from work package 3 (DEA or SFA, both if necessary). Unsupervised machine learning is used to classify the various Natura 2000 areas in order to reduce the complexity of the efficiency estimates. This allows the heterogeneity of the areas to be included in the estimation of individual farm efficiency. Using the farm and field data collected in work package 2, comparative farms inside and outside Natura 2000 areas can be identified (e.g. by matching) for a counterfactual analysis of the policy effects of Natura 2000 (e.g. using staggered DiD to take into account different points in time of the treatment). Using suitable weighting methods (e.g. raking), the individual policy effects are also extrapolated to the Austrian population if possible.
Arbeitspaket 5: Publikation und Dissemination
Dieter Kömle, Yvonne Stickler, Christoph Tribl, Sebastian Lakner
The findings will be summarised in a report for the BML. In addition, work package 5 will summarise the methods and results in the form of a scientific article. A submission will be made to a suitable scientific journal. By participating in international conferences and internal feedback rounds, the content and analysis are iteratively revised and improved until they are ready for publication. After publication, the results are presented to a broader public via suitable popular science formats. The methodological and content-related insights gained will be used to design follow-up projects.
Work 2025
The project is initiated in June 2025 with all project partners (kick-off meeting). This will be followed by the contextualisation of Natura 2000 (WP1) and data acquisition (WP2) in the fourth quarter.
Status of the project
The project is initiated for 2025.
Timetable
Project start: 06/2025
Project end: 05/2028
Literature
Ahlvik, Lassi, und Sebastiaan van Kooten. 2024. „Distributional Impacts of Conservation on Land Prices: Evidence From Natura 2000“. SSRN Scholarly Paper. Rochester, NY: Social Science Research Network. https://doi.org/10.2139/ssrn.4938652.
Bach, Philipp, Malte S. Kurz, Victor Chernozhukov, Martin Spindler, und Sven Klaassen. 2024. „DoubleML: An Object-Oriented Implementation of Double Machine Learning in R“. Journal of Statistical Software 108 (Februar):1–56. https://doi.org/10.18637/jss.v108.i03.
Bogetoft, Peter, und Lars Otto. 2010. Benchmarking with DEA, SFA, and R. Springer Science & Business Media.
Brümmer, Bernhard, und Jens-Peter Loy. 2000. „The Technical Efficiency Impact of Farm Credit Programmes: A Case Study of Northern Germany“. Journal of Agricultural Economics 51 (3): 405–18. https://doi.org/10.1111/j.1477-9552.2000.tb01239.x.
Callaway, Brantly, und Pedro H. C. Sant’Anna. 2021. „Difference-in-Differences with Multiple Time Periods“. Journal of Econometrics, Themed Issue: Treatment Effect 1, 225 (2): 200–230. https://doi.org/10.1016/j.jeconom.2020.12.001.
Chernozhukov, Victor, Denis Chetverikov, Mert Demirer, Esther Duflo, Christian Hansen, Whitney Newey, und James Robins. 2018. „Double/debiased machine learning for treatment and structural parameters“. The Econometrics Journal 21 (1): C1–68. https://doi.org/10.1111/ectj.12097.
Coelli, Timothy J., Dodla Sai Prasada Rao, Christopher J. O’Donnell, und George Edward Battese. 2005. An Introduction to Efficiency and Productivity Analysis. Springer Science & Business Media.
Cunningham, Scott. 2021. Causal Inference: The Mixtape. Yale University Press.
Djokoto, Justice. 2015. „Technical Efficiency of Organic Agriculture: A Quantitative Review“. Studies in Agricultural Economics 117 (2): 61–71. https://doi.org/10.7896/j.1512.
European Commission. 2018. Farming for Natura 2000: Guidance on How to Support Natura 2000 Farming Systems to Achieve Conservation Objectives, Based on Member States Good Practice Experiences. LU: Publications Office. https://data.europa.eu/doi/10.2779/85823.
Getzner, Michael, und Michael Jungmeier. 2002. „Conservation policy and the regional economy: the regional economic impact of Natura 2000 conservation sites in Austria“. Journal for Nature Conservation 10 (1): 25–34. https://doi.org/10.1078/1617-1381-00003.
Goodman-Bacon, Andrew. 2021. „Difference-in-differences with variation in treatment timing“. Journal of Econometrics, Themed Issue: Treatment Effect 1, 225 (2): 254–77. https://doi.org/10.1016/j.jeconom.2021.03.014.
Grupp, Tristan, Prakash Mishra, Mathias Reynaert, und Arthur A. van Benthem. 2023. „An evaluation of protected area policies in the European Union“. 31934. NBER Working Paper Series. USA: National Bureau of Economic Research. https://www.nber.org/system/files/working_papers/w31934/w31934.pdf.
Imbens, Guido, und Keisuke Hirano. 2004. „The Propensity Score with Continuous Treatments“. In Applied Bayesian MOdelling and Causal Inference from Missing Data Perspectives, herausgegeben von Andrew Gelman und Xiao-Li Meng. Wiley.
Koemle, Dieter, Sebastian Lakner, und Xiaohua Yu. 2019. „The impact of Natura 2000 designation on agricultural land rents in Germany“. Land Use Policy 87 (September):104013. https://doi.org/10.1016/j.landusepol.2019.05.032.
Kumbhakar, Subal C., und C. A. Knox Lovell. 2003. Stochastic Frontier Analysis. Cambridge University Press.
Lakner, Sebastian, Yves Zinngrebe, und Dieter Koemle. 2020. „Combining Management Plans and Payment Schemes for Targeted Grassland Conservation within the Habitats Directive in Saxony, Eastern Germany“. Land Use Policy 97 (September):104642. https://doi.org/10.1016/j.landusepol.2020.104642.
Latruffe, Laure, Boris E. Bravo-Ureta, Alain Carpentier, Yann Desjeux, und Víctor H. Moreira. 2017. „Subsidies and Technical Efficiency in Agriculture: Evidence from European Dairy Farms“. American Journal of Agricultural Economics 99 (3): 783–99. https://doi.org/10.1093/ajae/aaw077.
Mennig, Philipp, und Johannes Sauer. 2020. „The impact of agri-environment schemes on farm productivity: a DID-matching approach“. European Review of Agricultural Economics 47 (3): 1045–93. https://doi.org/10.1093/erae/jbz006.
Nowak, Anna, Tomasz Kijek, und Katarzyna Domańska. 2015. „Technical Efficiency and Its Determinants in the European Union“. Agricultural Economics (Zemědělská Ekonomika) 61 (6): 275–83. https://doi.org/10.17221/200/2014-AGRICECON.
Stetter, Christian, Philipp Mennig, und Johannes Sauer. 2022. „Using Machine Learning to Identify Heterogeneous Impacts of Agri-Environment Schemes in the EU: A Case Study“. European Review of Agricultural Economics 49 (4): 723–59. https://doi.org/10.1093/erae/jbab057.
Thiam, Abdourahmane, Boris E. Bravo-Ureta, und Teodoro E. Rivas. 2001. „Technical Efficiency in Developing Country Agriculture: A Meta-Analysis“. Agricultural Economics 25 (2–3): 235–43. https://doi.org/10.1111/j.1574-0862.2001.tb00204.x.
Thiermann, Insa, und Thomas Bittmann. 2023. „Should I stay or should I go? The impact of nature reserves on the survival and growth of dairy farms“. Journal of Environmental Management 328 (Februar):116993. https://doi.org/10.1016/j.jenvman.2022.116993.
Tsiafouli, Maria A., Evangelia Apostolopoulou, Antonios D. Mazaris, Athanasios S. Kallimanis, Evangelia G. Drakou, und John D. Pantis. 2013. „Human Activities in Natura 2000 Sites: A Highly Diversified Conservation Network“. Environmental Management 51 (5): 1025–33. https://doi.org/10.1007/s00267-013-0036-6.
Uehleke, Reinhard, Heidi Leonhardt, und Silke Hüttel. 2024. „Counterfactual Evaluation of Two Austrian Agri-Environmental Schemes in 2014–2018“. Agricultural Economics 55 (1): 27–40. https://doi.org/10.1111/agec.12805.
Uthes, Sandra, und Bettina Matzdorf. 2012. „Studies on Agri-Environmental Measures: A Survey of the Literature“. Environmental Management 51 (1): 251–66. https://doi.org/10.1007/s00267-012-9959-6.
Translated with DeepL.com (free version)